AITeQ: a machine learning framework for Alzheimer's prediction using a distinctive five-gene signature

Area of Science:

• Genomics
• Computational Biology
• Neuroscience

Background:

• Neurodegenerative diseases like Alzheimer's present significant global health challenges.
• Complex etiology and lack of reliable biomarkers hinder early diagnosis.
• RNA-sequencing (RNA-seq) offers potential for biomarker discovery.

Purpose of the Study:

• To develop a machine learning (ML) model for identifying Alzheimer's disease using RNA-seq data.
• To create an accessible tool, Alzheimer's Identification Tool (AITeQ), for diagnostic support.
• To identify a minimal set of genes indicative of Alzheimer's disease.

Main Methods:

• Analysis of RNA-seq data to identify differentially expressed genes.
• Application of ML protocols including feature selection, model training, and hyperparameter tuning.
• Development of AITeQ using an optimized ensemble ML model (logistic regression, naive Bayes, support vector machine).

Main Results:

• Identified 87 differentially expressed genes from RNA-seq data.
• Reduced the gene set to five key genes (CNKSR1, EPHA2, CLSPN, OLFML3, TARBP1) through feature selection.
• Developed AITeQ, an ensemble ML model demonstrating high performance in identifying Alzheimer's disease.

Conclusions:

• AITeQ provides a novel, data-driven approach for Alzheimer's disease identification.
• The identified gene set and ML model show promise for early diagnostic applications.
• The AITeQ tool is publicly available for further research and development.